server,ui: Add debugging for quiesced ranges

I suspect that cockroachdb#26257 is caused by the unquiescedReplicas map
introduced in cockroachdb#24956 getting out of sync with the per-replica
quiescent flag. Add debug pages to help us see if that's happening.

Release note: None

pkg/server/problem_ranges.go

@@ -125,12 +125,17 @@ func (s *statusServer) ProblemRanges(
 					problems.NoLeaseRangeIDs =
 						append(problems.NoLeaseRangeIDs, info.State.Desc.RangeID)
 				}
+				if info.Problems.QuiescentEqualsTicking {
+					problems.QuiescentEqualsTickingRangeIDs =
+						append(problems.QuiescentEqualsTickingRangeIDs, info.State.Desc.RangeID)
+				}
 			}
 			sort.Sort(roachpb.RangeIDSlice(problems.UnavailableRangeIDs))
 			sort.Sort(roachpb.RangeIDSlice(problems.RaftLeaderNotLeaseHolderRangeIDs))
 			sort.Sort(roachpb.RangeIDSlice(problems.NoRaftLeaderRangeIDs))
 			sort.Sort(roachpb.RangeIDSlice(problems.NoLeaseRangeIDs))
 			sort.Sort(roachpb.RangeIDSlice(problems.UnderreplicatedRangeIDs))
+			sort.Sort(roachpb.RangeIDSlice(problems.QuiescentEqualsTickingRangeIDs))
 			response.ProblemsByNodeID[resp.nodeID] = problems
 		case <-ctx.Done():
 			return nil, status.Errorf(codes.DeadlineExceeded, ctx.Err().Error())

pkg/server/serverpb/status.proto

@@ -134,6 +134,12 @@ message RangeProblems {
   bool no_raft_leader = 3;
   bool underreplicated = 4;
   bool no_lease = 5;
+
+  // Quiescent ranges do not tick by definition, but we track this in
+  // two different ways and suspect that they're getting out of sync.
+  // If the replica's quiescent flag doesn't agree with the store's
+  // list of replicas that are ticking, warn about it.
+  bool quiescent_equals_ticking = 6;
 }
 
 message RangeStatistics {
@@ -179,6 +185,7 @@ message RangeInfo {
   CommandQueueMetrics cmd_q_global = 12 [ (gogoproto.nullable) = false ];
   storage.LeaseStatus lease_status = 13 [ (gogoproto.nullable) = false ];
   bool quiescent = 14;
+  bool ticking = 15;
 }
 
 message RangesRequest {
@@ -505,6 +512,11 @@ message ProblemRangesResponse {
       (gogoproto.casttype) =
           "github.com/cockroachdb/cockroach/pkg/roachpb.RangeID"
     ];
+    repeated int64 quiescent_equals_ticking_range_ids = 7 [
+      (gogoproto.customname) = "QuiescentEqualsTickingRangeIDs",
+      (gogoproto.casttype) =
+          "github.com/cockroachdb/cockroach/pkg/roachpb.RangeID"
+    ];
   }
   reserved 1 to 7;
   // NodeID is the node that submitted all the requests.

pkg/server/status.go

@@ -1094,16 +1094,18 @@ func (s *statusServer) Ranges(
 				WritesPerSecond:  rep.WritesPerSecond(),
 			},
 			Problems: serverpb.RangeProblems{
-				Unavailable:          metrics.Unavailable,
-				LeaderNotLeaseHolder: metrics.Leader && metrics.LeaseValid && !metrics.Leaseholder,
-				NoRaftLeader:         !storage.HasRaftLeader(raftStatus) && !metrics.Quiescent,
-				Underreplicated:      metrics.Underreplicated,
-				NoLease:              metrics.Leader && !metrics.LeaseValid && !metrics.Quiescent,
+				Unavailable:            metrics.Unavailable,
+				LeaderNotLeaseHolder:   metrics.Leader && metrics.LeaseValid && !metrics.Leaseholder,
+				NoRaftLeader:           !storage.HasRaftLeader(raftStatus) && !metrics.Quiescent,
+				Underreplicated:        metrics.Underreplicated,
+				NoLease:                metrics.Leader && !metrics.LeaseValid && !metrics.Quiescent,
+				QuiescentEqualsTicking: metrics.Quiescent == metrics.Ticking,
 			},
 			CmdQLocal:   serverpb.CommandQueueMetrics(metrics.CmdQMetricsLocal),
 			CmdQGlobal:  serverpb.CommandQueueMetrics(metrics.CmdQMetricsGlobal),
 			LeaseStatus: metrics.LeaseStatus,
 			Quiescent:   metrics.Quiescent,
+			Ticking:     metrics.Ticking,
 		}
 	}
 

pkg/storage/replica.go

@@ -6021,7 +6021,13 @@ type ReplicaMetrics struct {
 	Leaseholder bool
 	LeaseType   roachpb.LeaseType
 	LeaseStatus LeaseStatus
-	Quiescent   bool
+
+	// Quiescent indicates whether the replica believes itself to be quiesced.
+	Quiescent bool
+	// Ticking indicates whether the store is ticking the replica. It should be
+	// the opposite of Quiescent.
+	Ticking bool
+
 	// Is this the replica which collects per-range metrics? This is done either
 	// on the leader or, if there is no leader, on the largest live replica ID.
 	RangeCounter      bool
@@ -6050,6 +6056,10 @@ func (r *Replica) Metrics(
 	r.cmdQMu.Unlock()
 	r.mu.RUnlock()
 
+	r.store.unquiescedReplicas.Lock()
+	_, ticking := r.store.unquiescedReplicas.m[r.RangeID]
+	r.store.unquiescedReplicas.Unlock()
+
 	return calcReplicaMetrics(
 		ctx,
 		now,
@@ -6060,6 +6070,7 @@ func (r *Replica) Metrics(
 		leaseStatus,
 		r.store.StoreID(),
 		quiescent,
+		ticking,
 		cmdQMetricsLocal,
 		cmdQMetricsGlobal,
 	)
@@ -6084,6 +6095,7 @@ func calcReplicaMetrics(
 	leaseStatus LeaseStatus,
 	storeID roachpb.StoreID,
 	quiescent bool,
+	ticking bool,
 	cmdQMetricsLocal CommandQueueMetrics,
 	cmdQMetricsGlobal CommandQueueMetrics,
 ) ReplicaMetrics {
@@ -6099,6 +6111,7 @@ func calcReplicaMetrics(
 	m.Leaseholder = m.LeaseValid && leaseOwner
 	m.Leader = isRaftLeader(raftStatus)
 	m.Quiescent = quiescent
+	m.Ticking = ticking
 
 	// We compute an estimated range count across the cluster by counting the
 	// first live replica in each descriptor. Note that the first live replica is

pkg/storage/replica_test.go

@@ -8672,7 +8672,7 @@ func TestReplicaMetrics(t *testing.T) {
 			metrics := calcReplicaMetrics(
 				context.Background(), hlc.Timestamp{}, config.SystemConfig{},
 				c.liveness, &c.desc, c.raftStatus, LeaseStatus{},
-				c.storeID, c.expected.Quiescent, CommandQueueMetrics{}, CommandQueueMetrics{})
+				c.storeID, c.expected.Quiescent, !c.expected.Quiescent, CommandQueueMetrics{}, CommandQueueMetrics{})
 			if c.expected != metrics {
 				t.Fatalf("unexpected metrics:\n%s", pretty.Diff(c.expected, metrics))
 			}

pkg/ui/src/views/reports/containers/problemRanges/connectionsTable.tsx

@@ -38,14 +38,19 @@ const connectionTableColumns: ConnectionTableColumn[] = [
     title: "Underreplicated (or slow)",
     extract: (problem) => problem.underreplicated_range_ids.length,
   },
+  {
+    title: "Quiescent equals ticking",
+    extract: (problem) => problem.quiescent_equals_ticking_range_ids.length,
+  },
   {
     title: "Total",
     extract: (problem) => {
       return problem.unavailable_range_ids.length +
         problem.no_raft_leader_range_ids.length +
         problem.no_lease_range_ids.length +
         problem.raft_leader_not_lease_holder_range_ids.length +
-        problem.underreplicated_range_ids.length;
+        problem.underreplicated_range_ids.length +
+        problem.quiescent_equals_ticking_range_ids.length;
     },
   },
   { title: "Error", extract: (problem) => problem.error_message },

pkg/ui/src/views/reports/containers/problemRanges/index.tsx

@@ -168,6 +168,11 @@ class ProblemRanges extends React.Component<ProblemRangesProps, {}> {
           problems={problems}
           extract={(problem) => problem.underreplicated_range_ids}
         />
+        <ProblemRangeList
+          name="Quiescent equals ticking"
+          problems={problems}
+          extract={(problem) => problem.quiescent_equals_ticking_range_ids}
+        />
       </div>
     );
   }

pkg/ui/src/views/reports/containers/range/rangeTable.tsx

@@ -41,6 +41,7 @@ const rangeTableDisplayList: RangeTableRow[] = [
   { variable: "problems", display: "Problems", compareToLeader: true },
   { variable: "raftState", display: "Raft State", compareToLeader: false },
   { variable: "quiescent", display: "Quiescent", compareToLeader: true },
+  { variable: "ticking", display: "Ticking", compareToLeader: true },
   { variable: "leaseType", display: "Lease Type", compareToLeader: true },
   { variable: "leaseState", display: "Lease State", compareToLeader: true },
   { variable: "leaseHolder", display: "Lease Holder", compareToLeader: true },
@@ -204,6 +205,9 @@ export default class RangeTable extends React.Component<RangeTableProps, {}> {
     if (problems.unavailable) {
       results = _.concat(results, "Unavailable");
     }
+    if (problems.quiescent_equals_ticking) {
+      results = _.concat(results, "Quiescent equals ticking");
+    }
     if (awaitingGC) {
       results = _.concat(results, "Awaiting GC");
     }
@@ -449,6 +453,7 @@ export default class RangeTable extends React.Component<RangeTableProps, {}> {
         problems: this.contentProblems(info.problems, awaitingGC),
         raftState: raftState,
         quiescent: info.quiescent ? rangeTableQuiescent : rangeTableEmptyContent,
+        ticking: this.createContent(info.ticking.toString()),
         leaseState: leaseState,
         leaseHolder: this.createContent(
           Print.ReplicaID(rangeID, lease.replica),